摘要
SVM是数据挖掘中一种具有优良模式识别性能的新方法,该方法具有学习速度快、全局最优和泛化能力强等优点。首先利用支持向量机回归(SVR)构建辨识遥测不良数据的模型,在状态估计前通过比较预测值与实测值之间的差值来一次性辨识遥测不良数据。接着将状态估计后得到的标准残差作为支持向量机分类(SVC)的输入,依靠拓扑错误的残差特性来分类辨识出拓扑错误。通过对IEEE-30母线的仿真分析证明了该方法的有效性,现行状态估计器的效率及合格率可以得到很好的提高。
SVM is a new method with excellent pattern recognition properties in data mining, which has the advantages of fast learning, global optimum and high generalization. Firstly, support vector machine regression is utilized to establish the identi-fication model for the telemetric bad data, which compares the differences between the predicted values and the measured values before state estimation. Then the obtained standard residuals after state estimation are used as the input of SVC classification, and the topology error is identified based on the characteristics of these residuals. The efficiency of the proposed method is proven by the simulation analysis of IEEE -30 bus model, thus the efficiency and the percent of pass of the existing state estimators can be highly improved.
出处
《四川电力技术》
2013年第1期59-63,共5页
Sichuan Electric Power Technology
关键词
不良数据
电力系统状态估计
检测辨识
支持向量机
bad data
power system state estimation
detection and identification
support vector machine
